1. Technical Field
The present invention relates to the technical field of image processing apparatuses that detect a lane line projected as a subject, on the basis of a captured image that is obtained through imaging of the direction of travel of a vehicle with the apparatus.
2. Related Art
Vehicle systems that recognize a vehicle exterior environment include systems that recognize a subject on the basis of a captured image that is obtained through stereo-imaging of the direction of travel of a vehicle equipped the system by a pair of cameras. Systems which recognize thus a subject that is present in the direction of travel of the vehicle include systems that generate three-dimensional position information (information on the position in actual space) on the subject, including information on the distance up to the subject. In such systems, specifically, captured images of different viewpoints are obtained through stereo-imaging, the parallax of the subject is calculated on the basis of the captured images, and three-dimensional position information on the subject is generated on the basis of the parallax.
Recognition schemes of the vehicle exterior environment include schemes that detect lane lines that are formed on a road surface, such as white lane lines, and detect curves present in the direction of travel the gradient of the road surface, on the basis of a three-dimensional model of the lane lines.
Lane line detection is typically accomplished on the basis of luminance images, by exploiting a comparatively large difference between the road surface and white lane lines.
As used herein, the term “lane line” denotes a line that represents a boundary between traveling lanes, and that is formed by paint or the like, on the road surface.
Lane lines also include colored lane lines such as orange lane lines (yellow lane lines), other than white lane lines. The luminance of colored lane lines is lower than that of white lane lines. Accordingly, a lane line may not be detected, depending on the travel circumstances such as travel over a concrete road of comparatively high luminance, in which a sufficient luminance difference between the road surface and the lane line cannot be obtained, or the luminance of the lane line is lower than the road surface luminance.
Herein, detection based on a color difference image is effective in order to detect colored lane lines stably. For instance, it is effective to detect orange lane lines based on a V image (R-G) or a U image (B-G).
Although the precision of detection of colored lane lines can be enhanced through detection based on a color difference image, the detection precision of white lane lines in that case is lower than when the lane line is detected based on a luminance image. It would be accordingly conceivable to execute both detection based on a luminance image and detection based on a color difference image, for a shared image portion, in order to detect appropriately both a white lane line and a colored lane line (for example, see Japanese Patent No. 4365350).
However, performing two types of lane line detection for a shared image portion brings an increase in the processing load, which is undesirable.
The incidence of colored lane lines is usually smaller than that of white lane lines, and, accordingly, performing at all times lane line detection based on a color difference image together with lane line detection based on a luminance image also brings a undesirable increase in the processing load.